There is a workaround i found to this problem, if i have 500 rows, i have to execute them in 5 beeline/jdbc/hive cli sessions, so first session inserts 120 rows then another new session/new jdbc connection takes over and so on. I am not sure why this is happening at first place, where existing hive session is not able to see the hdfs file/directory names generated by inserts in that session and generates the exactly same file as for previous insert. ... View more